I took a very brief look. Many tests are failing with
assert_equals: wasClean should be true expected true but got false
This is happening on a normal browser-initiated close, so we shouldn't be failing. Either we have some bug that is triggered by the web-platform-tests but not by our layout tests, or there's something odd going on with the test harness.
There are some tests with different issues that might be easier to fix, eg. websockets/constructor/002.html.
ricea@, since you've taken a look and agree there are some things worth looking into here, should we perhaps mark this as 'Available' with an appropriate priority to get it off the "untriaged" Hotlist-Interop list? Blink>Network has a long tail of untriaged bugs :-(
I fixed up some w-p-t WebSocket tests recently: https://github.com/w3c/web-platform-tests/pull/7345. Unfortunately, there are many broken tests and we cannot justify resources to fix them. The good news is, in many cases the browsers were interoperable and only the tests were wrong.
This issue is very nearly "WontFix" but if some kind person fixes the tests then we can take another look.
WontFixing then, I think the right approach here would be for those incorrect tests to show up in a triage in upstream web-platform-tests, using up-to-date results from all browsers on wpt.fyi.
Comment 1 by ricea@chromium.org
, Oct 3 2016